Update libcontainer to 6460fd79667466d2d9ec03f77f3

Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
This commit is contained in:
Michael Crosby 2015-01-12 10:38:15 -08:00
parent cb4ea2c778
commit 115d8bac60
10 changed files with 395 additions and 40 deletions

View File

@ -68,7 +68,7 @@ if [ "$1" = '--go' ]; then
mv tmp-tar src/code.google.com/p/go/src/pkg/archive/tar
fi
clone git github.com/docker/libcontainer be02944484da197166020d6b3f08a19d7d7d244c
clone git github.com/docker/libcontainer 6460fd79667466d2d9ec03f77f319a241c58d40b
# see src/github.com/docker/libcontainer/update-vendor.sh which is the "source of truth" for libcontainer deps (just like this file)
rm -rf src/github.com/docker/libcontainer/vendor
eval "$(grep '^clone ' src/github.com/docker/libcontainer/update-vendor.sh | grep -v 'github.com/codegangsta/cli')"

View File

@ -105,7 +105,7 @@ func GetStats(systemPaths map[string]string) (*cgroups.Stats, error) {
stats := cgroups.NewStats()
for name, path := range systemPaths {
sys, ok := subsystems[name]
if !ok {
if !ok || !cgroups.PathExists(path) {
continue
}
if err := sys.GetStats(path, stats); err != nil {

View File

@ -174,7 +174,7 @@ func ParseCgroupFile(subsystem string, r io.Reader) (string, error) {
return "", NewNotFoundError(subsystem)
}
func pathExists(path string) bool {
func PathExists(path string) bool {
if _, err := os.Stat(path); err != nil {
return false
}
@ -183,7 +183,7 @@ func pathExists(path string) bool {
func EnterPid(cgroupPaths map[string]string, pid int) error {
for _, path := range cgroupPaths {
if pathExists(path) {
if PathExists(path) {
if err := ioutil.WriteFile(filepath.Join(path, "cgroup.procs"),
[]byte(strconv.Itoa(pid)), 0700); err != nil {
return err

View File

@ -120,6 +120,10 @@ type Config struct {
// Rlimits specifies the resource limits, such as max open files, to set in the container
// If Rlimits are not set, the container will inherit rlimits from the parent process
Rlimits []Rlimit `json:"rlimits,omitempty"`
// AdditionalGroups specifies the gids that should be added to supplementary groups
// in addition to those that the user belongs to.
AdditionalGroups []int `json:"additional_groups,omitempty"`
}
// Routes can be specified to create entries in the route table as the container is started

View File

@ -67,7 +67,7 @@ func TestIPCPrivate(t *testing.T) {
}
if actual := strings.Trim(buffers.Stdout.String(), "\n"); actual == l {
t.Fatalf("ipc link should be private to the conatiner but equals host %q %q", actual, l)
t.Fatalf("ipc link should be private to the container but equals host %q %q", actual, l)
}
}
@ -152,7 +152,7 @@ func TestIPCBadPath(t *testing.T) {
_, _, err = runContainer(config, "", "true")
if err == nil {
t.Fatal("container succeded with bad ipc path")
t.Fatal("container succeeded with bad ipc path")
}
}
@ -176,3 +176,34 @@ func TestRlimit(t *testing.T) {
t.Fatalf("expected rlimit to be 1024, got %s", limit)
}
}
func TestPIDNSPrivate(t *testing.T) {
if testing.Short() {
return
}
rootfs, err := newRootFs()
if err != nil {
t.Fatal(err)
}
defer remove(rootfs)
l, err := os.Readlink("/proc/1/ns/pid")
if err != nil {
t.Fatal(err)
}
config := newTemplateConfig(rootfs)
buffers, exitCode, err := runContainer(config, "", "readlink", "/proc/self/ns/pid")
if err != nil {
t.Fatal(err)
}
if exitCode != 0 {
t.Fatalf("exit code not 0. code %d stderr %q", exitCode, buffers.Stderr)
}
if actual := strings.Trim(buffers.Stdout.String(), "\n"); actual == l {
t.Fatalf("pid link should be private to the container but equals host %q %q", actual, l)
}
}

View File

@ -110,9 +110,44 @@ func Exec(container *libcontainer.Config, stdin io.Reader, stdout, stderr io.Wri
return -1, err
}
}
if !container.Namespaces.Contains(libcontainer.NEWPID) {
killAllPids(container)
}
return command.ProcessState.Sys().(syscall.WaitStatus).ExitStatus(), nil
}
// killAllPids itterates over all of the container's processes
// sending a SIGKILL to each process.
func killAllPids(container *libcontainer.Config) error {
var (
procs []*os.Process
freeze = fs.Freeze
getPids = fs.GetPids
)
if systemd.UseSystemd() {
freeze = systemd.Freeze
getPids = systemd.GetPids
}
freeze(container.Cgroups, cgroups.Frozen)
pids, err := getPids(container.Cgroups)
if err != nil {
return err
}
for _, pid := range pids {
// TODO: log err without aborting if we are unable to find
// a single PID
if p, err := os.FindProcess(pid); err == nil {
procs = append(procs, p)
p.Kill()
}
}
freeze(container.Cgroups, cgroups.Thawed)
for _, p := range procs {
p.Wait()
}
return err
}
// DefaultCreateCommand will return an exec.Cmd with the Cloneflags set to the proper namespaces
// defined on the container's configuration and use the current binary as the init with the
// args provided

View File

@ -170,7 +170,7 @@ func RestoreParentDeathSignal(old int) error {
}
// SetupUser changes the groups, gid, and uid for the user inside the container
func SetupUser(u string) error {
func SetupUser(container *libcontainer.Config) error {
// Set up defaults.
defaultExecUser := user.ExecUser{
Uid: syscall.Getuid(),
@ -188,12 +188,14 @@ func SetupUser(u string) error {
return err
}
execUser, err := user.GetExecUserPath(u, &defaultExecUser, passwdPath, groupPath)
execUser, err := user.GetExecUserPath(container.User, &defaultExecUser, passwdPath, groupPath)
if err != nil {
return fmt.Errorf("get supplementary groups %s", err)
}
if err := syscall.Setgroups(execUser.Sgids); err != nil {
suppGroups := append(execUser.Sgids, container.AdditionalGroups...)
if err := syscall.Setgroups(suppGroups); err != nil {
return fmt.Errorf("setgroups %s", err)
}
@ -273,7 +275,7 @@ func FinalizeNamespace(container *libcontainer.Config) error {
return fmt.Errorf("set keep caps %s", err)
}
if err := SetupUser(container.User); err != nil {
if err := SetupUser(container); err != nil {
return fmt.Errorf("setup user %s", err)
}

View File

@ -15,6 +15,8 @@
#include <unistd.h>
#include <getopt.h>
#define pr_perror(fmt, ...) fprintf(stderr, "nsenter: " fmt ": %m\n", ##__VA_ARGS__)
static const kBufSize = 256;
static const char *kNsEnter = "nsenter";
@ -22,6 +24,10 @@ void get_args(int *argc, char ***argv)
{
// Read argv
int fd = open("/proc/self/cmdline", O_RDONLY);
if (fd < 0) {
pr_perror("Unable to open /proc/self/cmdline");
exit(1);
}
// Read the whole commandline.
ssize_t contents_size = 0;
@ -34,6 +40,10 @@ void get_args(int *argc, char ***argv)
bytes_read =
read(fd, contents + contents_offset,
contents_size - contents_offset);
if (bytes_read < 0) {
pr_perror("Unable to read from /proc/self/cmdline");
exit(1);
}
contents_offset += bytes_read;
}
while (bytes_read > 0);
@ -91,8 +101,7 @@ void nsenter()
#ifdef PR_SET_CHILD_SUBREAPER
if (prctl(PR_SET_CHILD_SUBREAPER, 1, 0, 0, 0) == -1) {
fprintf(stderr, "nsenter: failed to set child subreaper: %s",
strerror(errno));
pr_perror("Failed to set child subreaper");
exit(1);
}
#endif
@ -124,9 +133,8 @@ void nsenter()
init_pid = strtol(init_pid_str, NULL, 10);
if ((init_pid == 0 && errno == EINVAL) || errno == ERANGE) {
fprintf(stderr,
"nsenter: Failed to parse PID from \"%s\" with output \"%d\" and error: \"%s\"\n",
init_pid_str, init_pid, strerror(errno));
pr_perror("Failed to parse PID from \"%s\" with output \"%d\"",
init_pid_str, init_pid);
print_usage();
exit(1);
}
@ -135,7 +143,7 @@ void nsenter()
argv += 3;
if (setsid() == -1) {
fprintf(stderr, "setsid failed. Error: %s\n", strerror(errno));
pr_perror("setsid failed");
exit(1);
}
// before we setns we need to dup the console
@ -143,9 +151,7 @@ void nsenter()
if (console != NULL) {
consolefd = open(console, O_RDWR);
if (consolefd < 0) {
fprintf(stderr,
"nsenter: failed to open console %s %s\n",
console, strerror(errno));
pr_perror("Failed to open console %s", console);
exit(1);
}
}
@ -154,51 +160,60 @@ void nsenter()
memset(ns_dir, 0, PATH_MAX);
snprintf(ns_dir, PATH_MAX - 1, "/proc/%d/ns/", init_pid);
int ns_dir_fd;
ns_dir_fd = open(ns_dir, O_RDONLY | O_DIRECTORY);
if (ns_dir_fd < 0) {
pr_perror("Unable to open %s", ns_dir);
exit(1);
}
char *namespaces[] = { "ipc", "uts", "net", "pid", "mnt" };
const int num = sizeof(namespaces) / sizeof(char *);
int i;
for (i = 0; i < num; i++) {
char buf[PATH_MAX];
memset(buf, 0, PATH_MAX);
snprintf(buf, PATH_MAX - 1, "%s%s", ns_dir, namespaces[i]);
int fd = open(buf, O_RDONLY);
if (fd == -1) {
// Ignore nonexistent namespaces.
// A zombie process has links on namespaces, but they can't be opened
struct stat st;
if (fstatat(ns_dir_fd, namespaces[i], &st, AT_SYMLINK_NOFOLLOW) == -1) {
if (errno == ENOENT)
continue;
pr_perror("Failed to stat ns file %s for ns %s",
ns_dir, namespaces[i]);
exit(1);
}
fprintf(stderr,
"nsenter: Failed to open ns file \"%s\" for ns \"%s\" with error: \"%s\"\n",
buf, namespaces[i], strerror(errno));
int fd = openat(ns_dir_fd, namespaces[i], O_RDONLY);
if (fd == -1) {
pr_perror("Failed to open ns file %s for ns %s",
ns_dir, namespaces[i]);
exit(1);
}
// Set the namespace.
if (setns(fd, 0) == -1) {
fprintf(stderr,
"nsenter: Failed to setns for \"%s\" with error: \"%s\"\n",
namespaces[i], strerror(errno));
pr_perror("Failed to setns for %s", namespaces[i]);
exit(1);
}
close(fd);
}
close(ns_dir_fd);
// We must fork to actually enter the PID namespace.
int child = fork();
if (child == -1) {
pr_perror("Unable to fork a process");
exit(1);
}
if (child == 0) {
if (consolefd != -1) {
if (dup2(consolefd, STDIN_FILENO) != 0) {
fprintf(stderr, "nsenter: failed to dup 0 %s\n",
strerror(errno));
pr_perror("Failed to dup 0");
exit(1);
}
if (dup2(consolefd, STDOUT_FILENO) != STDOUT_FILENO) {
fprintf(stderr, "nsenter: failed to dup 1 %s\n",
strerror(errno));
pr_perror("Failed to dup 1");
exit(1);
}
if (dup2(consolefd, STDERR_FILENO) != STDERR_FILENO) {
fprintf(stderr, "nsenter: failed to dup 2 %s\n",
strerror(errno));
pr_perror("Failed to dup 2\n");
exit(1);
}
}
@ -208,9 +223,7 @@ void nsenter()
// Parent, wait for the child.
int status = 0;
if (waitpid(child, &status, 0) == -1) {
fprintf(stderr,
"nsenter: Failed to waitpid with error: \"%s\"\n",
strerror(errno));
pr_perror("nsenter: Failed to waitpid with error");
exit(1);
}
// Forward the child's exit code or re-send its death signal.

View File

@ -0,0 +1,70 @@
package nsenter
import (
"fmt"
"os"
"os/exec"
"os/signal"
"strings"
"syscall"
"testing"
)
func TestNsenterAlivePid(t *testing.T) {
args := []string{"nsenter-exec", "--nspid", fmt.Sprintf("%d", os.Getpid())}
cmd := &exec.Cmd{
Path: os.Args[0],
Args: args,
}
err := cmd.Run()
if err != nil {
t.Fatal("nsenter exits with a non-zero exit status")
}
}
func TestNsenterInvalidPid(t *testing.T) {
args := []string{"nsenter-exec", "--nspid", "-1"}
cmd := &exec.Cmd{
Path: os.Args[0],
Args: args,
}
err := cmd.Run()
if err == nil {
t.Fatal("nsenter exits with a zero exit status")
}
}
func TestNsenterDeadPid(t *testing.T) {
c := make(chan os.Signal)
signal.Notify(c, syscall.SIGCHLD)
dead_cmd := exec.Command("true")
if err := dead_cmd.Start(); err != nil {
t.Fatal(err)
}
defer dead_cmd.Wait()
<-c // dead_cmd is zombie
args := []string{"nsenter-exec", "--nspid", fmt.Sprintf("%d", dead_cmd.Process.Pid)}
cmd := &exec.Cmd{
Path: os.Args[0],
Args: args,
}
err := cmd.Run()
if err == nil {
t.Fatal("nsenter exits with a zero exit status")
}
}
func init() {
if strings.HasPrefix(os.Args[0], "nsenter-") {
os.Exit(0)
}
return
}

View File

@ -0,0 +1,200 @@
{
"capabilities": [
"CHOWN",
"DAC_OVERRIDE",
"FOWNER",
"MKNOD",
"NET_RAW",
"SETGID",
"SETUID",
"SETFCAP",
"SETPCAP",
"NET_BIND_SERVICE",
"SYS_CHROOT",
"KILL"
],
"cgroups": {
"allowed_devices": [
{
"cgroup_permissions": "m",
"major_number": -1,
"minor_number": -1,
"type": 99
},
{
"cgroup_permissions": "m",
"major_number": -1,
"minor_number": -1,
"type": 98
},
{
"cgroup_permissions": "rwm",
"major_number": 5,
"minor_number": 1,
"path": "/dev/console",
"type": 99
},
{
"cgroup_permissions": "rwm",
"major_number": 4,
"path": "/dev/tty0",
"type": 99
},
{
"cgroup_permissions": "rwm",
"major_number": 4,
"minor_number": 1,
"path": "/dev/tty1",
"type": 99
},
{
"cgroup_permissions": "rwm",
"major_number": 136,
"minor_number": -1,
"type": 99
},
{
"cgroup_permissions": "rwm",
"major_number": 5,
"minor_number": 2,
"type": 99
},
{
"cgroup_permissions": "rwm",
"major_number": 10,
"minor_number": 200,
"type": 99
},
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 1,
"minor_number": 3,
"path": "/dev/null",
"type": 99
},
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 1,
"minor_number": 5,
"path": "/dev/zero",
"type": 99
},
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 1,
"minor_number": 7,
"path": "/dev/full",
"type": 99
},
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 5,
"path": "/dev/tty",
"type": 99
},
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 1,
"minor_number": 9,
"path": "/dev/urandom",
"type": 99
},
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 1,
"minor_number": 8,
"path": "/dev/random",
"type": 99
}
],
"name": "docker-koye",
"parent": "docker"
},
"restrict_sys": true,
"mount_config": {
"device_nodes": [
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 1,
"minor_number": 3,
"path": "/dev/null",
"type": 99
},
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 1,
"minor_number": 5,
"path": "/dev/zero",
"type": 99
},
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 1,
"minor_number": 7,
"path": "/dev/full",
"type": 99
},
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 5,
"path": "/dev/tty",
"type": 99
},
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 1,
"minor_number": 9,
"path": "/dev/urandom",
"type": 99
},
{
"cgroup_permissions": "rwm",
"file_mode": 438,
"major_number": 1,
"minor_number": 8,
"path": "/dev/random",
"type": 99
}
],
"mounts": [
{
"type": "tmpfs",
"destination": "/tmp"
}
]
},
"environment": [
"HOME=/",
"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
"HOSTNAME=koye",
"TERM=xterm"
],
"hostname": "koye",
"namespaces": [
{"type": "NEWIPC"},
{"type": "NEWNET"},
{"type": "NEWNS"},
{"type": "NEWUTS"}
],
"networks": [
{
"address": "127.0.0.1/0",
"gateway": "localhost",
"mtu": 1500,
"type": "loopback"
}
],
"tty": true,
"user": "daemon"
}